基于HMM的远动联调语音多层次背景噪音滤除模型  

HMM Based Multilevel Background Noise Filtering Model for Telecontrol Combined Modulation Speech

作　　者：肖大军 徐遐龄 刘涛 陈艳明 李鑫 于文娟 XIAO Dajun;XU Xia Ling;LIU Tao;CHEN Yanming;LI Xin;YU Wen juan(Central China Branch of State Grid Corporation of China,Hubei 430070,China;Shanghai Xiaoi Robot Technology Co.,Ltd.,Shanghai 201800,China)

机构地区：[1]国家电网有限公司华中分部,湖北武汉430070 [2]上海智臻智能网络科技股份有限公司,上海201800

出　　处：《长江信息通信》2023年第10期70-72,共3页Changjiang Information & Communications

基　　金：国家电网有限公司科技项目资助(合同号:SGHZ0000DKJS2200235)主站与厂站远动数据智能调试研究及应用(项目编号:5214DK220016)。

摘　　要：构建基于HMM的远动联调语音多层次背景噪音滤除模型,多层次滤除远动联调语音背景噪音,确保远动联调工作顺利完成。在远动联调语音识别阶段,输入提取到远动联调语音信号MFCC特征,使用改进Viterbi递推解码算法,获取远动联调语音MFCC特征与参数库中HMM参数相对应的概率密度,将概率密度最大值在远动联调语音字库中对应的字当作有效远动联调语音识别结果,完成背景噪音与无效语音数据初步滤除;在此基础上,使用平移不变小波去噪方法深度滤除有效远动联调语音背景噪音,完成远动联调语音多层次背景噪音滤除。实验结果表明:该方法应用后,背景噪音滤除较为彻底,保证远动联调语音语义准确性较高。The HMM-based multi-level background noise filtering model of telecontrol joint debugging speech is constructed to filter the background noise of telecontrol joint debugging speech at multiple levels to ensure the successful completion of telecontrol joint debugging.In the telecontrol joint debugging speech recognition stage,the MFCC feature of the telecontrol joint debugging speech signal is input and extracted.The improved Viterbi recursive decoding algorithm is used to obtain the probability density corresponding to the MFCC feature of the telecontrol joint debugging speech and the HMM parameter in the parameter database.The word corresponding to the maximum probability density in the telecontrol joint debugging speech database is regarded as the effective telecontrol joint debugging speech recognition result,and the background noise and invalid voice data are initially filtered;On this basis,the translation invariant wavelet denoising method is used to deeply filter the background noise of the effective telecontrol joint debugging speech and complete the multi-level background noise filtering of the telecontrol joint debugging speech.The experimental results show that after the application of this method,the background noise is filtered thoroughly,and the semantic accuracy of telecontrol joint debugging is guaranteed to be high.

关 键 词：HMM模型 远动联调语音 多层次 背景噪音 噪音滤除 MFCC特征 

分 类 号：TN912[电子电信—通信与信息系统]